Get Started  ·  FIVE STEPS, ALL LOCAL

From Clone ToFirst Session.

The repos are private until the public beta in Q4 2026. The steps below are exactly what will unlock then, everything runs locally, no account, no API key, no credit card.

◉
SOURCE ACCESS, COMING SOON · Q4 2026

Anvaya CLI and Anvaya Mind are private repositories until the public beta. Join the early-access list and we’ll send you the clone-and-build instructions the moment they’re public.

No spam. Early access only. Unsubscribe anytime.✓ You’re on the list. We’ll email you at launch.Enter a valid email address.
01

Prerequisites

Rust 1.75 or newer, and Ollama running locally with at least one model pulled. Ollama is what lets Anvaya run without an API key.

TERMINAL, STEP 01
# Rust toolchain, https://rustup.rs
$rustc --version
rustc 1.75.0 or newer
# Pull a coding model
$ollama pull devstral:latest
02

Build the CLI

Anvaya is not on crates.io or Homebrew yet; beta testers build it from source. The release profile produces a single static binary (14–19 MB) with no runtime dependencies.

TERMINAL, STEP 02
$git clone <your-access-url>
$cd anvaya
$cargo build --release
# Binary at target/release/anv
03

Install & initialize

Put anv on your PATH, then initialize the project you want it to work in. anv init creates the .anvaya/ directory that holds config, sessions, and backups.

TERMINAL, STEP 03
$export PATH="$PWD/target/release:$PATH"
$cd ~/code/your-project
$anv init
✓ .anvaya/ initialized
04

First session

Launch the TUI, or skip it entirely and run headless. Keys 1-8 switch screens, Tab cycles, Ctrl+P opens the palette, Y/N approves.

TERMINAL, STEP 04
$anv
# or headless, for pipes and CI
$anv --no-tui --plan-only "what would change to add Redis caching?"
$anv --continue
05

Memory starts itself

Memory runs as a local daemon. It scaffolds under .anvaya/mind/, starts on your first connected session, and runs nine background workers that keep the graph consolidated.

TERMINAL, STEP 05
# no separate install step
$anv init
# first session spawns the daemon
$anv mind stats
◆ nodes · links · sessions · decay active

Already Using Claude Code?

Keep Your Tool. Add The Memory.

Anvaya ships an MCP memory server, so it attaches to any MCP-capable client without replacing it. Drop this in your project root and the memory becomes available as a set of tools inside the editor or agent you already use.

.mcp.jsonproject root
# .mcp.json, in your project root
{
  "mcpServers": {
    "anvaya-mind": {
      "command": "mind-mcp",
      "args": ["--project", "."]
    }
  }
}

That exposes fourteen memory tools, including mind_query, mind_search, mind_record_change, mind_get_node, mind_feedback, mind_stats, session start/end, thread and delta retrieval, proactive candidates, intent, retro-promotion and federation. mind_query returns a packed context under a token budget you set, defaulting to 4,096.